Re: Brainlike Computers, Learning From Experience

[Note:  This comment comes from a reader of Dave Farber’s IP List.  DLH]

From: Gregory Aharonian 
Date: Sunday, December 29, 2013
Subject: Re A friendly comment on you neural net article
To: Dave Farber <>


Sorry to disagree with John, but absolutely nothing in that article is new.  Neural networks, AI applications, have been steadily developing for over 30 years, many before all of those 700+ Stanford students were born.  Every application he discusses has been amply discussed in countless articles and patents.  Neural network coprocessors commercially available?  That was new in the 1990s.  Neural networks in consumer services?  Also new in the 1990s, though much is behind the scenes.  Here’s some AI.  Next time I type “Panama” into Google search, how about it asking me if I mean Panama – the country, Panama City – Florida, Panama – the banana disease, or Panama – the hat?  Which was also first developed in the 1990s, though has yet to be innovated in the Valley.

No, the only thing interesting about anything AI at Stanford or anywhere else in the Valley is that all concern about job destruction has been eradicated.  After all, “smart computers” raises the question – “smart as what”?  Mushrooms?  No. Bonobos?  No.  Humans?  Yes.  I will concede that eradication is new, something new to mention in the article.  But beyond that, there are multiple delusions in the tech world about innovation.  Much of it is rediscovery, which is one reason the vast majority of Electronics patents are invalid, reflecting neither novelty nor inventive step, and one reason the economy limps along on bubbles.  New doesn’t need bubbles nor the Fed.

Greg Aharonian
Internet Patent News Service

From: John Markoff 
Date: Sunday, December 29, 2013
Subject: A friendly comment on you neural net article
To: Dave Farber <>

Hi Dave,

A fair comment and one that I knew would be made by the technical community. The article appears now because 2014 is the year when Qualcomm will commercialize it  in semiconductor form and I thought that was very much worth mentioning. I think the case can be made the neuroprocessor will become a widely used coprocessor in the same way that the GPU is ubiquitous…

on the point that neural nets have been around forever. Yes, but. I know about the fraud detection applications, but Last year was the first year that neural nets — deep learning — began to be used extensively in consumer services — visual search and speech recognition as deployed by Google and MSFT…


On Dec 29, 2013, at 8:35 AM, Dave Farber <> wrote:

I read your article and I understand the limits of New York Times in technical detail but I fail to understand what is new with the approach that was talked about in the article. This is been around for many many years there’ve been endless experiments and demonstrations showing that it’s actually useful. In fact it is used often in the financial area.
In the 1960s early I even designed a machine that will do nural nets and I was not the first one who did that


Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s